Updating/improving tesseract-ocr

75 views
Skip to first unread message

Lorenza Romano

unread,
Aug 10, 2018, 1:26:29 PM8/10/18
to tesseract-ocr
Hi all

I'm a neophyte in Tesseract.

Currently I'm using tesseract 3.0.5, I'm using it with the Italian language and newspaper pages in pdf format.
I'm converting pdf to png (with Imagemagick) with  300 and 600 dpi, following all these suggestions https://github.com/tesseract-ocr/tesseract/wiki/ImproveQuality

It works quite well, but it makes some errors. So what I would like to do is improving OCR recognition adding some specific examples, i.e. the sentences in which it makes some errors.
Is it possible? A sort of active learning....

I went through this one


and I succeded in creating a new traineddata file.... but then I discovered that it is not possible updating the existing traineddata file for the Italian language...
One should make the entire training whit all the pngs and boxes files.

Is there a way of doing this sort of updating/improving? Maybe with Tesseract 4?

Thanks in advance
Lorenza

 
Reply all
Reply to author
Forward
0 new messages